View my account

Using D435 for tracking 3D color objects

Comments

3 comments

  • MartyG

    According to the Chief Technical Officer of the RealSense Group at Intel (agrunnet), an object should be trackable with D435 so long as its size takes up 9 pixels or greater on the image.

    https://github.com/IntelRealSense/librealsense/issues/4175#issuecomment-507448389 

    It is recommended that depth and color images are aligned.  This can help to separate the foreground from the background.  Also, it is recommended that the D435 model's optimal depth resolution of 848x480 is used.

     

    If applying an image tag to the containers is a possibility and you are able to use ROS in your project, a way to accomplish your goal quickly may be to use the aruco_ros package.  With this, you can detect the "pose" (position and rotation) of objects with a marker tag on them called an aruco tag

    https://github.com/pal-robotics/aruco_ros 

    0
    Comment actions Permalink
  • Venkata83

    Thanks for the quick reply Marty. What if the 3d cylinders are small of the order of 7mm in dia and 10 mm long. I am not sure I can put markers on them. Would it be possible to track the colored cylinders in real-time using something similar to color tracking in OpenCV ?

    0
    Comment actions Permalink
  • MartyG

    Yes, you can track by color too using a platform such as OpenCV.  Googling for realsense tracking color opencv pose may help to find some useful leads. 

    An example that I found with such a search in the link below uses RealSense D435, ROS and OpenCV to detect objects of a particular color and track their pose.

    Edit: Link no longer exists.  Here is a different RealSense-compatible ROS project for segmenting a ball by its color and detecting its position.

    https://github.com/ctu-mrs/object_detect

    0
    Comment actions Permalink

Please sign in to leave a comment.